PARIS As the new coronavirus spreads throughout the world, scientists are using the latest computer modeling techniques to predict its consequences: from the eventual number of cases and deaths to the peak of the outbreak.
In recent decades, the power of new computer processors is combined with increasingly sophisticated mathematical models to give health authorities a much better idea of how fast diseases are likely to spread.
Despite an inevitable margin of error, scientists can now predict the path that the new coronavirus will take.
France’s health minister, Agnes Buzyn, said Friday that she had had telephone conversations with her G7 counterparts to accelerate the modeling of the possible spread and severity of diseases.
The models are constructed in a composite manner, taking into account the data of the known history of the virus (transmission rates, mortality and recovery), as well as trends in human behavior, such as air traffic patterns.
But it is not enough to be sophisticated: the best models are also adaptable.
“It is a compensation: the more you refine the model, the harder it is to manipulate it and open the door to a wide margin of error,” said Arnaud Banos, of the National Center for Scientific Research in France.
To adapt in real time to outbreak developments, researchers conduct simulations by inserting new data as they arise.
“This could be the emergence of a new epidemic hot spot or a new public health measure that the model could not have anticipated,” Banos said.
A British team from the London School of Hygiene and Tropical Medicine used computer models last week to estimate that the peak of the outbreak at the epicenter of the virus, the Chinese city of Wuhan, could arrive in mid or late February.
“Of course, there is a lot of uncertainty about when this will happen and how big it could be,” said team member Adam Kucharski.
Rowland Kao, Professor of Veterinary Epidemiology and Data Science, University of Edinburgh, cautiously welcomed the results.
“This is an analysis by an experienced and talented team, but as always the limitations of the available data will affect your predictions,” he said.
A separate study published in early February had modeled that at least 75,000 people were infected in Wuhan.
Some models compiled at the beginning of the outbreaks yield erroneous results due to the relative lack of data on which to base them.
This was the case of the BSE outbreak in Britain in the 1990s.
“Some models issued by accredited research groups said there would be up to 136,000 cases,” said Arnaud Fontanet of the Institut Pasteur.
“These uncertainties were based largely on assumptions about the incubation period of the diseases.”
In total, 177 people contracted the disease.
But computer models have come a long way in 20 years and are increasingly benefiting from the contributions of artificial intelligence.
According to Banos, this allows scientists to detect what are called “weak signals” that could be crucial in determining the accuracy of the models.
“These could be individual exchanges on social networks about the symptoms,” he said.
“The idea is to permanently collect large amounts of data so that weak signals are automatically collected and related to the evolution of the disease,” Banos added.